Standard Paricles Improvments

[Oldcode]

Hi,

I'm trying to do a project where I need high viscosity. But this business of having to crank the Sub Steps up to ridiculous levels is really getting in the way.

Right now, I'm doing a project where I want the viscosity at 300. Now, from stuff I've read, 300 minimum Sub Steps should be enough, but I've got a shot with several shots where I still get the popping particles. I've raised the Sub Steps up to almost 400, and it's still happening. It's getting to the point where it's not worth the time for each frame to simulate.

Any possibility of improvements for standard particles in the next version? Any chance of Viscosity settings coming back to Hybrido?

Thanks,

[tsn]

To me, 300 substeps aren't that high, and values of 500, 600, or more are sometimes necessary to get the desired result. As always, it depends on your scene and what you're trying to achieve. Have you every thought of decreasing "Int Pressure"? Low values (<0.1) will counteract the fluid's tendency to expand. Another option is to use a "k Speed" daemon with "Limit & Keep" set to "Yes". Finally, a drag force daemon can help to avoid these exploding particles. I also found it often better to use fixed substeps instead of the adaptive method. I don't know your scene, but you could also add a "k Isolated" daemon to kill straying particles. The "Isolation time" should be set to sth. like 0.01 to delete them immediately.

The next major RF update will have a viscosity model for Hybrido. For SPH I can't tell you right now if there are any plans to update viscosity.

Cheers,
Thomas

[CG_Nema]

I've taken it up to 5000 and it went at a reasonable speed with command line simulations. GUI simulation was impossibly slow.

[Oldcode]

Hey tsn,

Thanks for the reply. Getting viscosity back in Hybrido is good news. Half the reason I bought 2013 was I was hoping to use Hybrido for this project for I'd heard of how much better it was going to be.

Right now, I'm doing a shot with about 650,000 standard particles with the viscosity at 300, and the min/max substeps at 400-425. That with some interactions with fairly complicated geometry makes simulation time around 18 minutes per frame on average. Ever since I bumped up to 400, it seems to be mostly behaving, but there are a few popping particles occasionally. I've got some script tools I use to remove stray particles which may get me through this. Maybe I'm just impatient compared to you.

My system is an AMD Phenom II 3.2Ghz with 6 cores. Not bad for a 3.5 year old system, but maybe very slow compared what you guys have. I tried using the Command Line, but as I reported in another thread here, it keeps telling me I don't have a valid license. Luis is perplexed because since the GUI is working, so should the Command Line.

Anyway, I am already using the K_Speed, which helps the fluid from sloshing around too much. I've got the mix/max set from 0-10. I've never used the Drag Force Demon, so that' something I could try. Any idea what setting I should use for thick viscous mud?

If I turn down the Internal Pressure, will that make my liquid take up less volume/space? If that happens, I'll have to add more particles to fill up the container I'm using which will make the simulation time go up.

Lastly, I'm thinking of biting the bullet and getting a new system. I may have the money for it in couple of months. I'm looking at an AMD 4.7Ghz with 8 cores. Can't afford Intel.

[CG_Nema]

You've got to figure out that command line license problem, it might solve your problem completely.

[tsn]

The thing with high viscosity is that is can be considered the fluid's inner friction and this means that you actually have more or less the same problem as with particle friction. The forces between the particles get very high, and this results in a higher acceleration, and finally in, well, ridiculous velocities. With higher substeps you're able to increase accuracy, in particular in those areas where stability problems occur. And that's also why fixed substeps are better.

18 mins/frame is definitely a long time, but for 650 k particles and complex geometry, I'd say that's not bad.

With regards your scripts: are you running them under "FramesPre" or "StepsPre"? From my experience, most of these scripts only make sense when they're run on a per-step basis when you're working with high viscosity. A few minutes of a frame's simulation time are also used by the script to run through the particles. With daemons you can get the same result, but faster. I also sometimes use a "Surface tension" daemon with "Balanced" set to "Yes" to keep the particles together and "Strength" value around 50.

For thick mud, values around 75 (-100) should already work.

You're not alone with the CL problem. I'm having similar problems here on my machines as well with the RFRK and the Job Manager. We're investigating on this issue, but it takes some time to find out what's really going there in the background. So, all I can say right now is:"Sorry for the inconvenience."

[tsn]

One more idea that came into mind right now: the Hot 'n' Cold plugin

http://www.realflow.com/resources/item/show/c/104

It uses a different viscosity model and replaces RF's internal method. It can be used to create highly viscous fluids and doesn't suffer from RF's stability problems. Unfortunately, it's only available for RF2012 at the moment, but since there aren't any changes in the SPH solver, it should be enough to recompile the source code with RF's 2013 SDK lib... I'll ask the developer to provide a new version.

Cheers,
Thomas

[Oldcode]

Hey Thomas,

Thanks again for the reply. The only script I'm really using is a very simple 1 line script that Otuama made for me to save changes I've made to a frame cache after I use the particle selection tool to delete stray particles.

One thing you did mention that caught my eye is the particle friction. In Particle Fluid Interaction, I've got particle friction cranked way up (1.0) to help the mud stick to the object. Could this be way I'm still getting popping particles even with high sub steps? The popping particles look like they are coming from the surface and the inside of my geometry.

I'll try your idea for the Surface Tension Demon. Now, I've got surface tension on the emitter set to 100. Is that too high? I'll also look into the hot and cold demon.

Thanks again.

[tsn]

The particle friction is indeed an important factor and something I forgot to mention in my last reply. With particle friction, the problems are actually the same as with high viscosity and a value of 1.0 is definitely high. You could also try using a moderate sticky value (the object has to static in this case) or a bounded, rather strong, drag force daemon instead or in conjunction with a particle friction of 0.5, for example.

A surface tension of 100 isn't too high in my opinion, but I recommend using the daemon, because of its balanced function.

[Oldcode]

Thanks again!

I've reduced the Particle Friction to 0.5 and I think it's already made a big difference. I might even set it to 0.25 or even 0.1. Those values are a still a lot bigger than the default. I learned to set the particles friction up high from a tutorial about high viscosity fluids, but I think I over did it. I'm trying to do a shot with a person falling into some mud and getting all dirty. I'll try some experiments and see how low I can go so I don't have to use such high substeps.

One last question, if I do use the Surface Tension Demon. Should I leave the Surface Tension setting for the emitter at 0?

Thanks,